Routledge Handbook of Theoretical and Experimental Sign Language Research
Chapter on Implicatures: Theoretical and Experimental Perspectives
Kathryn Davidson (Harvard University)
1. Formal pragmatics and the theory of implicature
The study of natural language meaning revolves around the problem of how one person can
communicate their thoughts to another successfully, even when there are in principle
infinitely many thoughts, many of which the speakers will have never attempted to previously
communicate. To do this successfully, two people engaged in a conversation must understand
the structure of their language: how the underlying syntactic structure supports semantic
combinations to build meaning for new sentences that speakers may have never heard before.
Understanding how this works is the domain of lexical semantics (word level) and
compositional semantics (phrase/sentence level). In addition to the logical meaning, however,
two people attempting to communicate new thoughts must also know something beyond the
literal, logical, meaning of a sentence. They must know what a given sentence means in a
particular context, because most sentences have different meanings in different contexts - this
is the domain of pragmatics. The problem of how language users apply contextual
information to literal meanings to finally arrive at the intended meaning, i.e. the
semantics/pragmatics interface, is one of the most challenging problems not just in linguistics,
but in cognitive science more broadly. This chapter is concerned with one of the most wellstudied phenomena at the semantics/pragmatics interface, known as implicature.
The term implicature was first used by the philosopher H. Paul Grice (1970), who was
interested in the semantics/pragmatics interface problem: what part of the meaning of a
sentence is hard-coded, and what part is context dependent? This is a particularly important
problem for formal semanticists, following in the tradition of Montague (1973) and his
followers tying formal semantics to syntax, and who analyze natural language using an
underlying system of formal logic. To see why, consider the example of the English word or:
in a formal logic class, or is usually translated using the disjunctive operator (symbolized ∨).
In this system, A ∨ B is true if A is true, or if B is true, or if both A and B are true. However,
in natural language, many people take She drank coffee or tea to mean that she either drank
coffee, or she drank tea, but she didn’t drink both. There is a tension, therefore, between the
most natural logical translation of some terms in natural language and their actual
interpretation by users of languages (in this case, whether or rules out the possibility of both).
Grice’s solution to this problem was to keep the meaning of natural language expressions as
close as possible to their logical equivalents. In doing this, he argued that much of linguistic
meaning comes about through pragmatic reasoning about the intentions of speaker, after and
above the literal, logical meaning of a sentence has been computed. He suggested that the way
interlocutors reason about each others’ language use can be characterized by four maxims of
conversation that users expect each other to be following:
1. The maxim of quantity: Be as informative as possible: gives as much information as
is needed, and no more.
2. The maxim of quality, Be truthful, do not give information that is false or that is not
supported by evidence.
3. The maxim of relation, Be relevant, say things that are pertinent to the discussion.
4. The maxim of manner, Be as clear, as brief, and as orderly as one can in what one
says, and avoid obscurity and ambiguity.
In the case of or, speakers try to be as informative as possible (following the maxim of
quantity) while still being truthful (following the maxim of quality). Grice suggested that if a
speaker already know both of the things that or connects to be true (e.g. that Mary drank
coffee, and also that Mary drank tea), then they should use a more informative term than or:
they should have used and (i.e. by saying Mary drank coffee and Mary drank tea). Since any
English speaker that uses or always has and also at their disposal, then usually when they say
or, they aren’t able to say and without violating the maxim of quality, and so generally or is
interpreted pragmatically as A or B but not both. Thus, or has a basic, logical meaning (where
one or the other or both disjuncts can be true, equivalent to ∨) as well as a pragmatically
enriched meaning (where one or the other disjunct can be true, but not both). The difference
between these meanings, Grice suggested, is due to a scalar implicature: an unsaid implication
that use of a weaker term implies that use of the stronger term would have been false.
It’s possible to see already that Grice’s (and many others’) view of this classic problem at the
semantic/pragmatic interface was English/Indo-European-centric: not all languages have
translation equivalents to English and and English or. The Yuman language Maricopa
famously uses a simple juxtaposition of noun phrases to mean either conjunction (1a) or
disjunction (1b) (Gil 1991), and a similar morpheme or construction used for both meanings
has also been reported for Japanese (Ohori 2004), American Sign Language (Davidson 2013)
(2), Warlpiri (Bowler 2014), and Cheyenne (Murray to appear).
(1)
a. John-s
Bill-s
v?aawuum.
John-NOM Bill- NOM 3.COME.PL.FUT
‘John and Bill will come.’
b. John-s
Bill-s
v?aawuumsaa.
John- NOM Bill- NOM 3. COME.PL.FUT.INFER
‘John or Bill will come.’
(2)
a. Context: I know that everyone had exactly one drink at the party. I wonder what
Mary chose to drink at the party, and ask my friend. She replies:
IXMary DRINK TEA COORD COFFEE.
‘She (Mary) drank tea or coffee.’
b. Context: I see that Mary looks very caffeinated and sick! I wonder how much
caffeine she may have had, and ask my friend. She replies:
IXMary DRINK TEA COORD COFFEE!
‘She (Mary) drank tea and coffee!’
In fact, the pattern seen in (2) in American Sign Language is available with multiple possible
forms of syntactic coordination, both coordination on the hands (Fig 1a) and coordination
through placement in space (Fig 1b). Either of these strategies can be used both for
conjunction (“and”) and disjunction (“or”) in the example in (2), and can be used for
coordinating noun phrases, verb phrases, and even full clauses (Davidson 2013).
Figure 1a: COORD-L
Figure 1b: COORD-shift
The pattern posed by ASL and other languages where conjunction and disjunction can be
pronounced in the same way raises new questions for a Gricean theory of implicature: if a
language conveys two separate concepts without the use of two contrasting lexical items, do
scalar implicatures still arise? In other words, are users of these languages more likely to get
the very “logical” meaning of “or”, that one or the other or maybe both disjunctions is true,
when they mean disjunction, given that there isn’t a stronger contrasting term and? And
further, what should be the underlying meaning (the semantics) for these kinds of items: are
they just ambiguous and happen to be pronounced the same, or is there a single shared
meaning, such that what seems like disjunction or conjunction comes about through the
pragmatics? This is important because users of these languages might arrive at a logical
meaning for or not because of a lack of implicatures, but just because the same term can also
mean and. In what follows, we will take ASL as a case study to see how users of ASL
actually interpret these constructions.
Before moving on, though, we’ll discuss one further note about scalar implicatures. If all of
this discussion were about the pair of logical words <or, and>, we might consider all of this
theoretical machinery about maxims, etc. to be overkill. In fact, though, the same kind of
reasoning can be applied to a very wide range of terms that stand in the same entailment
relationship to each other (meaning that use of one term is strictly stronger than another, in
the way that logical and is strictly stronger than logical or). Some other examples are positive
quantifiers <some, all> (3)-(4), negative quantifiers <few, no>, numbers <one, two, three,
four, …>, temperatures <warm, hot>, verbs <like, love>, and modals <can, must>.
(3)
(4)
Mary drank some of the tea. (English)
Implicates: Mary did not drink all of the tea.
IXMary DRINK SOME TEA
(ASL)
Implicates: Mary did not drink all of the tea.
Consider (5)-(7). One signature of implicatures in general, and scalar implicatures in
particular, is that they can be canceled (5a)-(7a) and strengthened (5b)-(7b).
(5)
(6)
(7)
a. Mary drank coffee or Mary drank tea, but not both.
b. Mary drank coffee or Mary drank tea, and maybe both.
a. Mary drank some of the tea, but not all.
b. Mary drank some of the tea, and maybe all.
a. Mary likes tea, but doesn’t love it.
b. Mary likes tea, and maybe even loves it.
This kind of strengthening and canceling is traditionally a marker of setting off the pragmatic
inferences of implicatures from entailments, which are part of the semantic/logical meaning
of a sentence. Consider (8)-(10), where the scalar items (<or, and>, <some, all>, <like, love>)
from (5)-(7) are reversed. Here, the cancellations and strengthenings are nonsensical because
the stronger term entails the weaker one, so you can’t defeat (a) or strengthen (b) the result.
Thus, the relationship between items in a scale is asymmetrical.
(8)
(9)
(10)
a. #Mary drank coffee and Mary drank tea, but not either.
b. #Mary drank coffee and Mary drank tea, and maybe either.
a. #Mary drank all of the tea, but not some.
b. #Mary drank all of the tea, and maybe some.
a. #Mary loves tea, but doesn’t like it.
b. #Mary loves tea, and maybe even likes it.
Traditionally, the difference between the canceling and strengthening ability of the
implicature and the inability to do so in entailments served to place the former squarely in the
domain of post-compositional pragmatics, and the latter in the domain of semantics. Recent
work, however, has raised the possibility of a more grammatical type of implicature, still able
to be canceled and strengthened, but triggered by something in the syntax/semantics of the
structure, not just in post-processing pragmatic reasoning. Chierchia, Fox, and Spector (2009)
point out some examples of what they suggest are embedded scalar implicatures, as in (11).
(11)
John believes that Mary drank coffee or tea.
Implicates: John believes that Mary drank coffee or tea but not both.
In cases like (11), Chierchia et al. argue that the strengthened meaning of or cannot have
arisen simply as a result of the processes of reasoning using Grice’s maxims because such
maxims should apply to entire utterances, after their logical meaning is computed. In (11), the
“stronger” version one must compare to is something smaller than the entire utterance, in fact
is the clause embedded within the matrix clause, under believe. They take this to suggest that
the strengthened version of or cannot always simply arise via two interlocutors following
conversational maxims about quantity, but rather that the language must also include
operators in the syntax that act specifically to do the work of the implicature in the formal
semantics. That is, they suggest that what was thought to be reasoning about how people use
language (pragmatic phenomena) are sometimes actually arising from the (literal, logical)
language itself, through the presence of what amounts to a “silent only” operator, also known
as an exhaustivity operator, making (11) logically equivalent to (11’).
(11’)
John believes that Mary drank [only] coffee OR tea.
This kind of argumentation has a number of consequences. For example, we saw earlier that
scalar implicatures differ from entailments in that they are cancelable. So, what is it that
causes a scalar implicature to arise, and why it is optional? A grammatical theory of
implicature must deal with this problem, and so is obligated to postulate linguistic triggers of
implicatures. One way to do this is to say that if a particular word belongs to a scale, then the
fact that it has scalar alternatives means that the exhaustivity operator can apply. So, a word
like some combines with the exhaustivity operator in a meaningful way because some has the
scalar alternative all, while other words like Mary or drink does not (at least, in a basic
context without further information). This raises new questions: does the shape of alternatives
matter? Do they have to have separate articulations, or just possible separate meanings?
Sign languages have the potential to provide answers to some of these questions. We saw
above that in ASL, conjunction and disjunction can be expressed with the same COORD
(COORD-shift or COORD-L). This raises the question of whether they are ambiguous, i.e. and
and or just happen to be pronounced alike in ASL (like the English financial and river bank).
Another possibility is that they are two translations of a single underlying lexical item that has
a more general meaning (something like “collect these into a set”). A third possibility is that
another trigger, such as nonmanual marking (movements of the face, torso, etc. not including
the hands) provides the distinction, and if this is the case, raises the question of whether
nonmanual marking can distinguish two items for the purposes of a scale. Davidson (2013)
provides theoretical arguments to rule out ambiguity, but the latter two possibilities remain.
The following sections will look at experimental work investigating these issues by looking at
the interpretation of scalar words in American Sign Language. Sign languages can shed light
on these important questions about scalar implicatures in a number of ways, including the
grammatical versus pragmatic debate, as well as what kinds of lexical items can form
alternatives for each other. Within the field of sign language research, experimental
investigation of scalar implicatures can also help uncover the ways that the modality of
language communication influences pragmatic inferences that are drawn, which is currently a
very open question. For example, does the use of space in general, and spatial loci in
particular, influence inferences about the maxim of quantity? We turn to these issues next.
2. Experimental investigations of implicatures
In spoken languages, scalar implicatures are one of the most experimentally well-studied
phenomena at the interface of semantics and pragmatics, for a number of good of reasons.
First, it was discovered that children behave differently than adults when it comes to
interpretations of scalar items: they seem to interpret sentences more “logically”, i.e. with a
weak reading (some as some and maybe all) without the strengthening provided by the scalar
implicature (some as some and not all). (Noveck 2001; Papafragou and Musolino 2003;
Huang and Snedeker 2009, a.o.). Within sign languages, it is still an open question whether
young children learning ASL behave in the same way, although given that the pattern holds in
many languages, we would expect that it would be. (There has been some investigation on the
acquisition of scalar implicatures by deaf adults who are late learners of a sign language,
compared to deaf adults who are native signers, which we will return to later.) A second
reason that scalar implicatures have been studied so extensively using experiments within
spoken languages is that even adults’ interpretation can be manipulated by a variety of
factors, including cognitive load (De Neys and Schaeken 2007) and linguistic and
extralinguistic context (Bott and Noveck 2004; Grodner, et al. 2010, a.o.), and so provide an
informative area for understanding the relationship between cognition and language meaning.
A third reason, important from the perspective of a theoretical linguist, is that the
interpretations given by speakers in carefully controlled contexts can help determine the right
choice among competing theories of the grammar of scalar implicature. Since Chierchia
(2006), Fox (2007), and Chierchia, Fox, and Spector (2008) argued for the possibility that
scalar implicatures are grammatical, there has been an explosion of research to determine how
adults’ interpretations vary, and whether they form an integrated part of sentence meaning in
the same way as other grammatical/compositional parts.
Given the high interest in the psycholinguistics of scalar implicatures within experimental
linguistics, it is therefore natural to ask whether the same occurs in sign languages, and
whether the sign language modality can help answer any questions that have been raised by
experimentalists and theoreticians. So far, American Sign Language seems to have been the
only sign language with experimental work on scalar implicature, and this has only been
studied in adults in a single series of three studies: one focused on modality (Davidson 2014)
which we focus on in the next section; one on the issue of alternatives with conjunction and
disjunction (Davidson 2013) that will be the focus of the following section; and a third, which
we turn to finally, on an acquisition study of late learners that begins to extend the study of
scalar implicatures in children to questions about the effects of late first language acquisition
among deaf adults (Davidson and Mayberry 2015).
3. Scalar implicatures in the sign modality
A Gricean theory view of scalar implicatures makes universal predictions about natural
language pragmatics: in any context in which a language has terms that can be strictly ordered
through entailment (e.g. <some, all>, <or, and>, etc.), then use of a weaker term will
implicate that use of the stronger term would violate the maxim of quality (i.e. would be
false). This should be how any participants in a discussion expect interpretation to proceed if
both are following the maxims. Because these maxims are not specific to the spoken language
modality, it is reasonable to expect to discover similar behavior in sign languages. There is,
however, very little work on the pragmatics of the visual/manual use of space for
communication, whether in gesture with spoken languages or in sign languages, and so it is
quite possible that there are semantic/pragmatic phenomena involved specifically with the use
of space that may be uncovered in work on scalar implicatures in sign languages.
In a first study to address the issue of scalar implicatures in a sign language, Davidson (2014)
compared English native speakers’ interpretations of sentences in English to ASL native
signers’ interpretations of sentences in ASL in a computer-based behavioral study. Three
types of scales were compared across the two languages in this study. The first two,
quantifiers and numerals, are two of the most well studied scales in English, and so
counterparts in ASL acted as controls, and were tested in a way that made no use of threedimensional space. These were hypothesized to lead to similar interpretations between the
English and the ASL groups. A third scale was tested that would rely more on the
manual/visual modality: a list of items was said in order in English (e.g. “There is a candle
and a globe”), while the same items were listed and assigned to locations in space in ASL
using referential loci (“There is a candle at point a and a globe at point b”). Ad hoc scales can
be constructed from lists of items, so that if in fact three items were present (e.g. a wallet, a
candle, and a globe), but the target sentence only mentioned two of them, then that sentence
was being underinformative, similar to using “some” to describe a situation when all is true.
Investigating any differences between English speakers’ and ASL signers’ interpretations of
such underinformative descriptions could help shed light on the pragmatics of assigning
spatial loci in ASL. Controls were created for each underinformative test case: these controls
involved a match between description and picture (“Match controls”), or a Mismatch (Figure
2). Each trial in the experiment consisted of a picture and a video (with a native speaker of
English or a native signer of ASL, depending on the participant group), and participants were
simply asked whether they were “satisfied that the picture matched”, in which case they
should press the “smile” button, otherwise press the “frown” button (inset, Figure 2).
!
Figure 2: A scheme of the experimental design for the study in Davidson (2014). The numbers (“4”) indicate
how many of each trial type were seen by each participant; trials were counterbalanced using the Latin Square
method such that each situation (colored cans, number of bears, etc.) only appeared in one trial type per
participant. Inset shows a screenshot of what participants viewed during a single (quantifier) trial.
The results of this study showed that, as expected, the Quantifier and Number scales elicited
scalar implicatures (rejections of underinformative descriptions) in both English and in ASL,
with no significant differences between the two groups on either scale. On the Ad Hoc scale,
however, where the ASL sentences made use of 3-dimensional space in assigning items to
loci, there were (marginally) more rejections by signers in ASL than by speakers of English in
underinformative examples. In other words, when the stimulus stated that there was a candle
and a globe (and in ASL assigned these to locations in space), signers in ASL were more
likely to reject this description when there was also a wallet than English speakers were. On
all control items, English speakers and ASL signers behaved similarly (Table 1).
Table 1: Mean rejection rates for each sentence type in English and in ASL from Davidson (2014), with
standard deviations following in parentheses.
English
ASL
Quantifiers
Numbers
Ad Hoc
Quantifiers
Numbers
Ad Hoc
Match
(Control)
0
0
0
0.19 (0.18)
0
0.09 (0.19)
Mismatch
(Control)
1
1
1
1
1
1
Underinformative
(Test)
0.77 (0.38)
0.96 (0.14)
0.54 (0.45)
0.84 (0.23)
1
0.88 (0.35)
This initial study of scalar implicatures in the sign language modality served two purposes.
First, it shows that whatever principles of conversation and language use, such as Grice’s
Maxims, lead English speakers to reject underinformative descriptions on quantifiers and
numbers, the same behavior is found among ASL signers, so they are no indications that these
principles differ between spoken and sign languages. Second, there are some indications that
when space can be used in ASL, it may have pragmatic consequences. This raises many more
questions than answers, and first among these is whether co-speech gesture accompanying
English sentences would have the same result. Hopefully future research can determine
whether this is a property of linguistic loci in particular, or a more general pragmatic use that
is also found in co-speech gesture. Another question this research raises is what happens
when a scale looks very different between ASL and English in a way that is unrelated to
space? For some answers to this question, we’ll turn in the next section to experimental work
on the interpretation of conjunction and disjunction in ASL.
4. Scalar implicatures based on conjunction/disjunction in ASL
Sometimes sign languages can inform natural language linguistics not just because they occur
in a different modality, but simply because they provided a rich new source of understudied
languages. As we noted above, English has separate words for conjunction (“and”) and
disjunction (“or”), but other languages are able to use the same syntactic coordinator to
express either of these meanings. This is the case for at least a few understudied spoken
languages, as well as American Sign Language (ASL). Davidson (2014) uses this property of
ASL to understand the role of scalar alternatives in implicature calculation: given that <or,
and> is one of the most well-studied scales in English (along with quantifiers and numbers,
which we just saw behave similarly in ASL and in English), we would expect similar
implicature patterns in ASL under normal conditions. However, because ASL has what
Davidson calls “general use coordinators” that simply perform syntactic coordination but do
not specify the logical relationship of either disjunction or conjunction, it is quite possible that
the pattern of implicatures looks very different for the coordination scale in ASL.
To test whether this linguistic difference leads to different patterns of implicature, a group of
native speakers of English and a group of native signers of ASL (a different group from the
previous study) participated in an experiment using the same methods described above to
study modality and scalar implicatures. The only difference this time was the scales
compared: quantifiers <some, all> were still a baseline, but now they were compared with
coordination. In English, coordination was instantiated with <or, and>, while in ASL, the
same concepts were conveyed through nonmanually distinguished COORD-shift (Figure 3).
CUPa COORD-SHIFT(AND) BOWLb
a.
CUPa COORD-SHIFT(OR) BOWLb
b.
Figure 3: Nonmanual differences between conjunctive interpretations of COORD-shift (3a) and disjunctive
interpretations (3b).
The results of this study were surprising and robust: while both groups (English speakers and
ASL signers) treated quantifiers in the same way, calculating scalar implicatures in the crucial
underinformative test trials, they differed in their treatment of the coordination scale. English
speakers calculated scalar implicatures for coordinators <or, and> just like they did for
quantifiers <some, all>, both at the same rate of 77%. However, ASL signers, who robustly
calculated scalar implicatures for quantifiers <SOME, ALL> (at a rate of 90%), did not do so
for coordinators, which they only rejected as underinformative (when COORD-shift(OR) was
used even though both disjuncts were true) at a rate of 35%. This is particularly surprising
given that they generally accepted the Match controls (at a rate of 80%) and rejected the
Mismatch controls (at a rate of 83%) for the coordination scale, which both also used
nonmanual markings, just like in the Test case, so the results could not have been obtained
just by ignoring nonmanuals. Taken together, this pattern suggests that while nonmanuals
were salient and used for purposes of truth conditions, they did not contribute enough of a
contrast to form a pragmatic scale for purposes of scalar implicature calculation. We can
conclude, then, that a scale must be formed of two separate lexical items, and that other
information (nonmanuals, context, etc.) is not enough to provide contrast for the purpose of
creating scalar alternatives. Along with other properties of these coordinations, these results
suggest a formal semantic analysis of general use coordination that involves a single
underspecified semantics for conjunctive and disjunctive interpretations (Davidson 2013).
Table 2: Mean rejection rates for each sentence type in English and in ASL from Davidson (2013), with
standard deviations following in parentheses.
English
ASL
Quantifiers
Coordination
Quantifiers
Coordination
Match
(Control)
0
0.02 ( 0.07)
0.05 (0.16)
0.20 (0.20)
Mismatch
(Control)
1
0.92 (0.16)
0.98 (0.08)
0.83 (0.17)
Underinformative
(Test)
0.77 (0.38)
0.77 (0.36)
0.90 (0.13)
0.35 (0.21)
The consequences of this study for the theory of scalar implicature are cross-linguistic: the
finding wasn’t primarily due to the modality of the language (a signed language), but rather
the fact that ASL is a language other than English that has this particular property of
coordination alternatives (ASL). Therefore, we would expect that other languages mentioned
that show the same use of general use coordinators would exhibit the same pattern. However,
it does seem to be the case that many sign languages make use of general use coordinators,
many of them in the same way, either a list buoy (Liddell 2003) (COORD-L) or simply
juxtaposing coordinates that are associated with referential loci (COORD-shift). Future
research would do well to explore whether the same patterns hold across other sign languages,
and how many use a form of coordination of this kind.
5. Acquisition of scalar implicatures: Theory
We’ve seen so far that adult native speakers of English and adult native signers of ASL show
similar rates of scalar implicature calculation in each language for a typical scale like
quantifiers. Although differences appear when space becomes involved (in ASL) or scales
have a different lexical structure (general use coordination), it seems clear that adults follow
the same process for calculation of typical scalar implicatures, in both sign and spoken
languages. This shouldn’t be taken for granted, as there are some groups who do not show the
same pattern of scalar implicature calculation, most notably young preschool-aged children.
One foundational study in scalar implicature acquisition was carried out by Papafragou and
Musolino (2003), who provided children with a truth/felicity judgment task. The experimenter
acted out a situation that involved toys in front of the children, and then described the scene;
the child participant then either accepted or rejected this description of the scene. In a crucial
trial example, an experimenter would show three horses, and all three would jump over a
fence. Then the experimenter would comment that “some of the horses jumped over the
fence.” Adults would overwhelming reject this description of such a situation (at a rate greater
than 90%), while children would overwhelmingly accept it (rejecting it less than 20% of the
time). The same pattern was seen with <start, finish>, while children would accept “start” as a
description of a situation that could also be described by “finish”, while adults would reject it.
On numbers (<two, three>), children were much more likely to reject an underinformative
description than in the other scales, though still to a lesser extend than adults.
In recent years, much work has been spent trying to understand why children’s behavior is so
different in this domain from adults, seemingly more “logical” (less pragmatic) in accepting
the literal meaning. One hypothesis is that children simply lack the cognitive processing
capacity to do this task in the way that it is presented. Chierchia et al. (2001) and Gualmini et
al. (2001) provided children with alternative descriptions, and asked which was better (e.g. for
above, “some” or “all” of the horses), and found that when explicitly given the choice
between a felicitious and an underinformative description, children performed much closer to
adult-like behavior. Katsos and Smith (2010) and Katsos and Bishop (2011) proposed instead
that children have a different tolerance for pragmatic infelicity than adults. They argue this
based on an experiment in which children are given more than three possible responses,
instead of just the usual binary accept/reject: they can reward a puppet with a small, medium,
or large fruit. These authors found that children frequently given a medium sized fruit to a
puppet who expresses a true but underinformative utterance, suggesting that children do
differentiate these both from false examples (for which they give a small fruit) and true and
felicitous examples (for which they give the best, large fruit). Finally, Barner et al. (2011)
argue that children have the appropriate cognitive capacity and do not have a different
tolerance, but merely cannot call up or do not know the appropriate scalar alternatives. They
illustrate this with children’s success on ad hoc trials, which are merely contextual scalar
alternatives (like the aforementioned <candle and globe, wallet and candle and globe>), and
for which children behave just like adults (Papafragou and Tantalou 2004; Stiller et al. 2011).
One problem facing studies of children’s behavior regarding semantics and pragmatics in
general, and scalar implicatures in particular, is that children’s cognitive and social
development is inextricably linked to their linguistic development. Just as they are developing
the cognitive maturity to hold and consider multiple possibilities at once, they are developing
the social norms of their communities that would teach them about appropriate pragmatic
tolerance, and they are also developing the linguistic knowledge of what alternatives form
scales in their language. It would be helpful, therefore, to be able to dissociated some of these
pieces of development from each other. This has been the goal behind recent work on scalar
implicatures in children and adults with autism, which has shown no differences between
adults with and without autism diagnoses – a surprise, given young children’s behavior and
the seemingly crucial role of pragmatics (Pijnacker et al. 2009, Chevallier et al. 2010). It has
also prompted recent studies on scalar implicatures in adult Broca’s aphasics, who have
experienced the cognitive and social development of neurotypical adults, but then impaired
language; this group also shows no significant impairments on scalar implicatures (Kennedy
et al. to appear). The next section focuses on yet another group where different aspects of
development are dissociated: Deaf adults with delayed first language exposure.
6. Scalar implicature and age of first language acquisition: Experiment
One way to dissociate language experience from cognitive and social maturity would be to
look at individuals who learned language at different ages. For example Slabakova (2010)
investigated scalar implicatures in adults’ learning a second language, namely Korean learners
of English. Finding, yet again, mostly success for this group of adults, they were nevertheless
not able to conclude whether language exposure was not important, because most scalar items
are very easily translated from one language to another. As we saw above, both ASL and
English show similar behavior with quantifiers so it would be easy for adult second language
learners to perform the task for most scales in one’s native language and simply translate back
for the purposes of the task. Therefore, the key would be to test individuals who learned a
language late in life, but did not have a first language to fall back on. Davidson and Mayberry
(2015) aimed to do just that through investigating scalar implicatures in Deaf adults whose
first language was all ASL, but who were exposed to ASL at varying ages.
Figure 4 illustrates the heterogeneity in the participants in the study in Davidson and
Mayberry (2015). As can be seen, participants included 8 native signers (who were exposed to
ASL from birth/age 0), 6 “early nonnative” signers who were first exposed to ASL as
children, and 7 “late nonnative” signers who were all exposed to ASL after puberty. Although
all participants, living in the United States, also had significant exposure to English, all selfreported that ASL was both their dominant and first language on a variety of measures.
Participants were tested in the same experimental paradigm as discussed above (and in fact,
the native signers were the same group of signers as reported in the coordination study). The
goal of the study was both to understand (a) the effect of delayed first language on scalar
implicature calculation and (b) an understanding of how delayed first language can effect
semantic/pragmatic competence, given the long line of research showing that deaf signers
exposed to a sign language at an earlier age outperform peers exposed to sign at later ages on
a wide variety of language measures both in sign languages and in subsequently learned
spoken languages (Mayberry et al. 2002, Boudreault & Mayberry 2006, a.o.). This study was
the first to target the level of semantics and pragmatics.
Figure 4: The age at time of testing, and age of first exposure to ASL, of the participants in Davidson and
Mayberry (2015)
Results showed that early native signers were not significantly different from nonnative
signers, so late nonnative signers (N=7) were compared to early (both native and nonnative)
signers a single group (N=14). These two groups generally performed similarly across a wide
variety of implicatures, with both showing more success than children typically do on such a
task. However, there was one area in which late learners did not calculate as many
implicatures as early/native signers, on the Test case in the quantifiers scale (recall, this is the
prototypical case of scalar implicature <some, all>). In other underinformative cases, like the
spatial ad hoc scale and the coordination scale, both of which had ASL specific properties
that could not have been transferred from English (the use of space and the underdetermined
<or, and> scale, respectively), the two groups showed no differences. This suggests that all
participants were indeed proficient and dominant in ASL, and yet there seems to be a slight
advantage for native signers to have early language input to have the strongest rates of scalar
implicature calculation in places that children also struggle with (e.g. scalar quantifiers).
Table 3: Mean rejection rates for each sentence type in ASL by each group of signers from Davidson and
Mayberry (2015), with standard deviations following in parentheses.
Early/Native Quantifiers
Learners
Spatial
Coordination
Late
Quantifiers
Learners
Spatial
Coordination
Match
(Control)
0.08 (0.18)
0
0.16 (0.18)
0.14 (0.20)
0.04 (0.09)
0.14 (0.13)
Mismatch
(Control)
0.95 (0.10)
1
0.88 (0.16)
1
1
0.79 (0.17)
Underinformative
(Test)
0.94 (0.11)
0.98 (0.06)
0.39 (0.26)
0.75 (0.29)
1
0.32 (0.19)
The conclusions from this study point toward some influence of the still developing language
component in children’s difficulty with scalar implicature calculation, but still, these adults
never looked like children in accepting a majority of underinformative descriptions. This
suggests that at least some part of children’s difficult does, then, lie in being children with
regards to cognitive and social development. With regard to modules of language affected by
delayed first language exposure, semantic/pragmatic effects seem to be present but limited,
and especially limited to areas where learned lexical knowledge (e.g. the existence of the
scalar contrast <some, all>) plays a crucial role; when deduction within context is enough
(e.g. in the spatial ad hoc case), the differences between groups disappear. Certainly more
work should be done to follow up and directly compare different levels of language exposure
and development, and age of participants taking the study, as well as more direct comparisons
to other atypical language development patterns, which can help illuminate both the
semantic/pragmatic skills of these groups further, and also understand what occurs in typical
language development both in sign and in speech.
7. Other implicatures based on modality
The focus of this chapter has so far been nearly all on one type of implicature, scalar
implicature, based on Grice’s Maxim of Quantity. However, sign languages are also
extremely fertile ground for testing and understanding more about conversational maxims
generally and other kinds of implicatures; in return, such pragmatic investigations show
immense promise for providing new light on longstanding puzzles in sign linguistics.
One example of an avenue for future research comes from the interaction of the language
modality with the Maxims of Manner and Quality. Davidson (2015) discusses a formal
semantics for classifier predicates, which along with a discrete morphological handshape can
encode analog/gradient information about the location and movement of objects (Emmorey
and Herzig 2003, Zucchi et al. 2012; see Zwitserlood 2012 for overview of classifiers). In
Davidson’s proposed semantics, the analog/gradient information is conveyed through a
demonstration, the level of detail of which is pragmatically determined. To take one example,
consider (12) in ASL, which involves each hand using an upright human classifier handshape
(CL-1), in a neutral location, moving toward each other. The interpretation is that two people
walked or went toward each other, and the default assumption is that it was a straight path,
but this is only implicated pragmatically. If in fact they moved in a zig-zag pattern, this can be
added information; the interlocutor need not deny the original statement (12a). However, if a
marked manner or path is used, as in (12b) where the path is given as zig-zag, then if the
interlocutor disagrees they must actually deny the original description (12b).
(12)
a. Right hand: CL-1(straight movement toward center)
Left hand: CL-1(straight movement toward center)
‘The two people walk toward each other.’
#Response: ‘No, they went toward each other in a zigzag pattern!’
#Response: ‘And actually, they went toward each other in a zigzag pattern!’
b. Right hand: CL-1(zigzag movement toward center)
Left hand: CL-1(zigzag movement toward center)
‘The individuals walked toward each other in a zigzag pattern.’
Response: ‘No, they went straight for each other!’
Much more detailed work should be done on the formal pragmatics of classifier constructions,
and what kinds of information is asserted, implicated, and presupposed; this is merely a
suggested beginning for an avenue for future research. New findings in this area would not
only break open new understanding about classifiers, but also may raise new questions about
how similar or different the inferential process is for their gradient spatial counterparts in cospeech gesture, helping to inform the longstanding puzzle of the relationship between sign,
speech, and gesture.
8. Conclusions
Nearly all of the studies discussed in this chapter form a single path of research, conducted
within communities using American Sign Language and English, and on a single type of
implicature (scalar implicatures). It is mostly because this is a newly growing field within
spoken languages, especially at the experimental level, that there remain a myriad of paths for
sign language research to follow up on this work both in scope (across languages and types of
implicatures) and in depth (with larger populations and more complex stimuli). The value of a
strong theory of semantics and pragmatics for a more general theory of natural language
should be clear: understanding how language works means understanding how it allows us to
communicate infinitely many thoughts with finite means.
Sign languages have been shown to have the same combinatorial finite means available to
them as spoken languages, and so the question is as relevant for sign as for speech: how does
the context contribute to the meaning of these parts? Moreover, sign languages may have
additional means that have yet to be studied to a significant degree in spoken languages, such
as the gradient/analog means of communication available in some parts of sign (e.g. the
movement and locations in classifier predicates) and the gestures that accompany speech;
understanding how these work in a three dimensional spatial system opens vast new areas of
research that show great promise for providing a more complete picture of language as a
communication system. Adding to this the new possibilities in language differences (e.g.
general use coordination in ASL) and in language acquisition patterns make studying
implicatures in sign languages an important contribution to natural language linguistics and
psycholinguistics. Finally, and most importantly, the semantic/pragmatic interface has
received significant attention by formal linguists only very recently, and has the potential to
provide important information to theorists as well as users and educators about the
development of language and meaning in Deaf signers with a variety of language
backgrounds. This level of linguistic analysis should not be ignored, and the tools of formal
pragmatics founded by Grice and developed over the decades since by many others can
provide a valuable source for new information about sign language and language
development.
(Word count: 6,975)
References
Barner, D., N. Brooks & A. Bale. (2011.) Accessing the unsaid: The role of scalar alternatives
in children’s pragmatic inference. Cognition 118(1). 84–93.
Bott, L., & Noveck, I. A. (2004). Some utterances are underinformative: The onset and time
course of scalar inferences. Journal of memory and language,51(3), 437-457.
Boudreault, P., & Mayberry, R. I. (2006). Grammatical processing in American Sign
Language: Age of first-language acquisition effects in relation to syntactic structure.
Language and Cognitive Processes, 21, 608–635.
Bowler, M. (2014). Conjunction and disjunction in a language without'and'. In Semantics and
Linguistic Theory (Vol. 24, pp. 137-155).
Chevallier, C., Wilson, D., Happé, F., & Noveck, I. (2010). Scalar inferences in autism
spectrum disorders. Journal of autism and developmental disorders, 40(9), 1104-1117.
Chierchia, G. (2006). Broaden your views: Implicatures of domain widening and the
“logicality” of language. Linguistic inquiry, 37(4), 535-590.
Chierchia, G., Fox, D., & Spector, B (2009). The grammatical view of scalar implicatures and
the relationship between semantics and pragmatics. Semantics: An international
handbook of natural language meaning, Berlin: Mouton de Gruyter.
Chierchia, G., S. Crain, M. T. Guasti, A. Gualmini & L. Meroni. (2001). The acquisition of
disjunction: Evidence for a grammatical view of scalar implicatures. In A. H.-J. Do, L.
Domínguez & A. Johansen (eds.), Proceedings of the 25th annual Boston University
Conference on Language Development [BUCLD 25], 157–168. Somerville, MA:
Cascadilla Press
Davidson, K. (2013). And’or ‘or’: General use coordination in ASL. Semantics and
Pragmatics, 6(4), 1-44.
Davidson, K. (2014). Scalar implicatures in a signed language. Sign Language &
Linguistics, 17(1), 1-19.
Davidson, K. (2015). Quotation, Demonstration, and Iconicity. To appear in Linguistics and
Philosophy.
Davidson, K., & Mayberry, R. I. (2015). Do Adults Show an Effect of Delayed First
Language Acquisition When Calculating Scalar Implicatures?. Language Acquisition,
(ahead-of-print), 1-26.
Emmorey, K. & Herzig, M. (2003). Categorical versus gradient properties of classifier
constructions in ASL. In K. Emmorey (Ed.), Perspectives on Classifier Constructions
in Signed Languages, Lawrence Erlbaum Associates, Mahwah NJ, pp. 222–246
Fox, D. (2007). Free Choice Disjunction and the Theory of Scalar Implicatures.
Presupposition and Implicature in Compositional Semantics. U. Sauerland and P.
Stateva. New York, Palgrave Macmillan: 71-120.
Gil, D. (1991). Aristotle goes to Arizona, and finds a language without and. Semantics
Universals and Universal Semantics, 96-130.
Grice, H. P. (1970). Logic and conversation (pp. 41-58).
Gualmini, A., S. Crain, L.Meroni, G. Chierchia & M. T. Guasti. (2001). At the
semantics/pragmatics interface in child language. Proceedings of Semantics and
Linguistic Theory XI. Ithaca, NY: CLC Publications, Cornell University.
Huang, Y. T., & Snedeker, J. (2009). Semantic meaning and pragmatic interpretation in 5year-olds: evidence from real-time spoken language comprehension. Developmental
psychology, 45(6), 1723.
Katsos, Napoleon & Dorothy V. M. Bishop. (2011). Pragmatic tolerance: Implications for the
acquisition of informativeness and implicature. Cognition 120(1). 67–81.
Katsos, N. & N.Smith. (2010). Pragmatic tolerance and speaker-comprehender asymmetries.
In K. Franich, K. M. Iserman & L. L. Keil (eds.), Proceedings of the 34th annual
Boston University Conference on Language Development [BUCLD 34], 221–232.
Somerville, MA: Cascadilla Press.
Kennedy, L., Romoli, J. , Bill, C., Folli, R., and Crain, S. (to appear) Presuppositions vs.
Scalar Implicatures: the view from Broca’s Aphasia. Proceedings of NELS 45, MIT,
Cambridge
Liddell, Scott. (2003). Grammar, gesture, and meaning in American Sign Language.
Cambridge: Cambridge University Press.
Mayberry, R. I., Lock, E., & Kazmi, H. (2002). Linguistic ability and early language
exposure. Nature, 417, 38.
Montague, R. (1973). The proper treatment of quantification in ordinary English (pp. 221242). Springer Netherlands.
Murray, S. E. (2013). Cheyenne Connectives.
De Neys, W., & Schaeken, W. (2007). When people are more logical under cognitive load:
Dual task impact on scalar implicature. Experimental psychology, 54(2), 128-133.
Grodner, D. J., Klein, N. M., Carbary, K. M., & Tanenhaus, M. K. (2010). “Some,” and
possibly all, scalar inferences are not delayed: Evidence for immediate pragmatic
enrichment. Cognition, 116(1), 42-55.
Noveck, I. A. (2001). When children are more logical than adults: Experimental
investigations of scalar implicature. Cognition, 78(2), 165-188.
Ohori, T. (2004). Coordination in mentalese. Typological Studies in Language, 58, 41-66.
Papafragou, A., & Musolino, J. (2003). Scalar implicatures: Experiments at the semantics–
pragmatics interface. Cognition, 86(3), 253-282.
Papafragou, A. & N. Tantalou. (2004). Children’s computation of implicatures. Language
Acquisition 12. 71–82.
Pijnacker, J., Hagoort, P., Buitelaar, J., Teunisse, J. P., & Geurts, B. (2009). Pragmatic
inferences in high-functioning adults with autism and Asperger syndrome. Journal of
autism and developmental disorders, 39(4), 607-618.
Slabakova, R. (2010). Scalar implicatures in second language acquisition. Lingua 120(10).
2444–2462.
Stiller, A. N. D. Goodman & M. C. Frank. (2011). Ad-hoc scalar implicature in adults and
children. Proceedings of the 33rd Annual Meeting of the Cognitive Science Society.
Zucchi, S., Cecchetto C., & Geraci, C. (2012). Event Descriptions and Classifier Predicates in
Sign Languages. Presentation FEAST in Venice, June 21, 2011.
Zwitserlood, I. (2012). Classifiers. In Sign language: An international handbook (pp. 158186). Mouton de Gruyter.